4 DISCUSSION
During the 2018/2019 influenza season in Japan, baloxavir was prescribed for 43% of children aged 7–15 years receiving antiviral treatment for influenza. In comparison with NAIs, baloxavir prescription was not associated with an increase in resource utilization within 9 days of treatment, with the exception of the comparison with oseltamivir.
With respect to baloxavir use and clinical consequences among children, data are limited to one published trial (including its secondary analyses), a single-arm study without an active comparator drug, and small case series reported from Japan [6,14, 19-22]. From a public health standpoint, it remains unclear whether baloxavir use and virus with reduced susceptibility to baloxavir would increase healthcare resource use; this is why we designed our study to focus on healthcare utilization rather than on the relationship between variant virus and its downstream clinical outcomes in patients.
As compared with NAIs, baloxavir prescription was not overall associated with increase in resource utilization within 9 days of treatment, except in one subgroup analysis comparing with oseltamivir (aOR = 1.011;P <.001). In this subgroup analysis, we also found that baloxavir reduced resource use relative to zanamivir (aOR = .995;P =.039). In these subgroup analyses, we did not perform P-value adjustment for multiple comparisons for three reasons. First, although sophisticated statistical approaches are proposed to account for multiple comparisons, it is still controversial when and how the adjustment for multiple comparisons should be done [23, 24]. Second, statistical testing without scientifically sound hypothesis may result in a false discovery that is difficult to explain biologically [24]. Our research hypothesis de novo was, due to less-susceptible virus, baloxavir use might result in the increased healthcare use as compared with all agents of NAI class, rather than a single NAI agent. Third, our large-scale data could detect a small but statistically significant difference that lacked clinical relevance [25]. Therefore, our subgroup analyses should be viewed as exploratory ones. However, 1% of excess (or decrease) in healthcare resource use may be substantial where baloxavir was prescribed for millions of patients in one year, such as in Japan. Future researches are accordingly warranted to assess whether our subgroup findings could be explained by chance alone or not. Until then, it should be noted that the effect of baloxavir was at best equivalent to that of oseltamivir [26].
In our study, oseltamivir accounted for about 15 % of antiviral prescriptions for influenza, which proportion may be lower than in other countries. There are two plausible reasons. First, Japan is the only country to date that approved laninamivir for treatment of influenza. This long-acting NAI with single-inhalation application is sometimes preferred over a 5-days course of oseltamivir because of its convenience [27]. Second reason is the concern for potential adverse event of oseltamivir including neuropsychiatric symptoms among teenagers [28], which negative association is currently considered unlikely [17, 29]. Aside from the low frequency of oseltamivir prescription, Japan is a top consumer of antivirals for influenza, which are commonly prescribed even for children at low risk [27]. The prescription pattern should be interpreted in this Japanese-specific context, but we believe that this issue does not affect our primary analysis, that is, the association between antiviral selection (ie, baloxavir vs NAIs) and the subsequent medical resource use.
We did not integrate return visits to healthcare providers into the primary composite outcome. In Japan, asymptomatic children recovering from influenza often revisit physicians to affirm whether they can return to school. Because this type of revisit could not be differentiated in the claims record from a visit due to continuing symptoms, return visits were only analyzed as one of the secondary outcomes. In our data, revisits occurred in 47% of children, and this event was independent of initial the antiviral class prescribed. Similarly, death was not considered an outcome because influenza-related death was expected to be infrequent among school-aged children.
This research describes the first-season experience of baloxavir use in routine care setting involving pediatric patients at various risk levels for severe influenza, which we believe to be the strength of our study. This study also has limitations. First, our observation was based on a single-year data with limited influenza B activity. Circulating strains and the proportion of drug-resistant virus differ year by year, and the further data are needed to ascertain whether our findings are reproducible. Second, there were a number of unmeasured factors because our data source was not constructed for research purposes; these unmeasured factors included symptom duration, the interval between symptom onset and drug use, the reason for selecting the antiviral drug prescribed, vaccination status of each children (an out-of-pocket service in Japan), and the indication of further medical resource use. Our findings might therefore be subject to bias from these effect-modifiers. Third, the generalizability of our research is unclear to other age group or other counties with different healthcare system. Fourth, whether patients adhered to the prescribed treatment is unknown. Zanamivir requires a longer treatment course than baloxavir (5 days vs. a single dose); hence, it is possible that the adherence rate was lower in this group, and this difference may contribute to higher odds of medical resource use in this group than in the baloxavir group. Finally, the accuracy of the administrative claims record is uncertain or debatable [30]. However, we believe that the exposure (defined by prescription record) and outcomes (ie, resource utilization such as hospitalization, tests and antibiotics use) were both recorded with high accuracy because of the incentive for reimbursement by the healthcare providers [31]. Further, comorbidities at high risk of severe influenza were all independently associated with future resource utilization, which may indicate that the impact of disease misclassification was minimal, if it occurred at all.
In summary, as compared with NAIs, baloxavir prescription was not overall associated with increases in healthcare resource utilization within 9 days of treatment, except in one exploratory comparison with oseltamivir. Future research efforts are warranted in broader clinical contexts; these involve the following seasons, other regions outside Japan, and the populations of different age group.